Friday, August 24, 2018

Preventing a GFCI from tripping on transmitted RF

Figure 1:
The installed GFCI (and chokes) in the
utility room/laundry room.
Click on the image for a larger version.
Several months ago I had need to rewire an outlet in my kitchen:  Rather than having just two outlets with a multi-outlet adapter I decided to replace it with a pair of duplex outlets to accommodate everything that would be plugged in.

It turned out that the electrical circuit for this outlet came from my (semi-unfinished) utility room where it powered the washing machine and gas clothes dryer.  Because it is recommended that both laundry and kitchen areas be equipped with some sort of protection against electrical shock, I decided to install a GFCI (Grouns Fault Circuit Interrupter) at that point (it's an older home) to protect both.  The installation wasn't all that difficult, the hardest bit was cutting away a bit of tile on the kitchen backsplash to allow the installation of a bigger box to house the two additional outlets.  In the semi-unfinished utility room I replaced what had been a single duplex outlet with that pictured in Figure 1.

Everything worked fine, except that when I transmitted on 40 meters, especially with the amplifier, I would find not only the outlet in the kitchen dead, but also the Ooma VOIP phone adapter, which was also plugged into the Utility room outlet served by this GFCI which showed as having been triggered.  Clearly, the GFCI was "seeing" some RF energy and "tripping out".

Why it happens:

The mere presence of RF at the GFCI isn't likely enough to cause it to trip - more likely, it was some RF current flowing through the GFCI - and the way that one typically wires a GFCI and uses it makes it more likely that this could happen.  Taking a look at Figure 2 helps to understand the situation.
Figure 2:
Interconnection of the GFCI to other circuits and loads.  As mentioned in the text, most of these connections are made via long cable runs - which will naturally act as antennas.  As can be seen from the drawing, the AC power from the circuit breaker (itself a long cable run) goes through the GFCI which then has wires that radiate out from there, acting as sort of a multi-wire dipole that permits RF current to flow through the GFCI.  The fact that these units are "grounded" means little at radio frequencies where about any length of wire can act as an antenna!
Click on the image for a larger version.

At first glance, this diagram shows nothing special - just normal interconnection.  But in terms of RF, there are a bunch of antennas here:
  • AC Power from the circuit breaker - This line runs from the middle of the house to a corner, where the circuit breaker panel is located:  A lot of RF can be intercepted and flow through here!
  • Interconnect to the kitchen - This goes up, through the floor and into the kitchen - yet another leg of the antenna.
  • Items plugged in in the kitchen outlet - These further extend the "antenna" of the wiring to the kitchen.
  • The washing machine - This is a large, metal box that can act as a sort of antenna.
  • The clothes dryer - Like this washing machine, this is a large box - and it has a metal gas pipe attached to it that more or less goes to ground.
  • The Ooma phone interface - There is an extension cord (also an antenna) that goes to this device, which is capacitively coupled (via isolation transformer windings) to both in-house phone wiring and the Ethernet cable connected to it.

The end result is that RF can flow through the GFCI, mimicking a current imbalance that can be detected by its internal circuitry as if it were a real current imbalance on the line and neutral wires, causing it to trip - and the susceptibility of it can depend on the amplitude and frequency of the RF energy.  As is the nature of anything that acts as antenna, the amount of RF energy that is picked up depends on many things, such as the length of the wire, where it is routed, what is connected to it, and the frequency of the RF energy and the transmitter power, all of these things working together in ways that are often inscrutible.

While keeping the RF outside the house (e.g. preventing it from coming into the shack by assuring that balanced feedlines are actually balanced and that coaxial cables are adequately decoupled to prevent current from flowing on their shields) is of great help - and highly recommended - but significant energy can still be picked up "over the air" by mains wiring.

Preventing false trips:

Regardless of the specific situation, the trick to preventing the GFCI from tripping out in the presence of RF is to keep RF from flowing through it!

Practically speaking, it is very difficult to fundamentally alter where wires go and what is connected to them as these are usually part of the wiring in the house, behind walls, making it inconvenient to change - and that assumes that one somehow knew exactly how RF pick-up might occur from an antenna onto these conductors.   The easiest way of preventing a problem is to keep RF from getting into the GFCI in the first place using series inductance to choke the RF currents.

To see what was done, consider Figure 3, which is an annotated version of Figure 2, above:
Figure 3:
An annotated version of Figure 2 showing the locations of the added ferrite devices.  Devices weren't placed on the cords to the washer or dryer as they were quite short compared to the other conductors - and because it wasn't required.
Click on the image for a larger version.
To figure out what I needed to do I first unplugged everything from the related outlets.  In my case, the GFCI no longer tripped out on RF so I started plugging things back in, noting when it started to trip out again:  If I'd already plugged something else in by this point I removed it to see if it still tripped, or if it simply contributed to the problem, tipping it over the edge.  In other words, you'll probably want to tray all possible combinations (within reason, of course) while noting the results.

Having done this, I determined that one main culprit was an extension cord to the Ooma device (see below) but it seemed that every appliance added just a bit to the problem.

The drawing above demonstrates several things:
    Figure 4:
    Three snap-on chokes on the
    wires going to the box with the
    Click on the image for
    a larger version.
  • Any long run (more than a couple of feet/meters) has on it a ferrite choke, located as close to the GFCI unit as possible.
By placing the choke as close as physically possible to the GFCI unit, there is less wire that can act as antenna.  This also reduced the amount of RF that can be coupled from one conductor into another from parallel runs of wire.
  • In my installation, the conductor that comes from the circuit breaker and that which goes to the kitchen outlet are paralleled for several feet/meters.  Near the GFCI one can see that both conductors go through a single ferrite for common-mode decoupling while a bit farther away, we see that there is a ferrite device on each, individual cable as shown in Figure 4.
In Figure 4 we see the three snap-on chokes:  The one with both cables is just above the junction box (bottom of the picture), the one on the yellow cable is just above it and near the top of the picture is the one on the black cable.  These were so-placed because there was just enough slack in the wire to allow their installation at those locations.
The choke just above the box provides common-mode suppression for signals that might appear on both cables while the individual chokes suppress the different signals that might appear on each.  This is probably overkill, but it seems to be sufficient.

  • There is a choke on the green extension cord (that goes to the Ooma box located a few feet away) that is located very near the plug (and GFCI). See Figure 5.
Figure 5
A multi-turn choke on the extension cord.
Click on the image for a larger version.
This choke actually consists of three turns in the core (as many as would fit!) to provide maximum reactance on this conductor.  It is this wire that seemed to be particularly problematic in coupling RF into the GFCI - possibly because this cord ran near some copper piping and furnace duct work, not to mention being connected to a device that has a degree of capacitive coupling into the house telephone wiring and, to a lesser extent, Ethernet cables.

I did not put similar chokes on the power cords connecting to the washing machine and clothes dryer, not finding it necessary to do so.  Being that they were right next to each other, they are likely to have very similar RF potential - and since we've effectively isolated all of the other RF paths to/from the GFCI with chokes, we've broken up possible paths for current flow, anyway.

Choke selection:

One of the problems with using snap-on ferrite chokes is that they are typically of little efficacy, the added reactance being quite small.  With the little reactance, lower frequencies - particularly 160-40 meters - may not be strongly attenuated.

The "typical" chokes that one finds on electronic devices are of a different material and have most of their efficacy at high HF and VHF/UHF frequencies where there is most concern when the device in question is facing EMC testing that may be required by regulatory agencies.  What we need is to use a ferrite device that is especially suited for lower HF frequencies.

Ferrite devices are available in a wide variety of "mixes" - the mix being a different formulation designed specifically to impede signals over certain ranges, the efficacy of most efficacy using being inversely related to the permeability of the mix:  In other words, materials with lower permeability often work adequately at higher frequencies while materials with high permeability are more effective at low frequencies - but often suffer somewhat at higher frequencies.

In general, on HF you are better off with the (less common, more expensive) higher permeability materials - and the higher the permeability, the better - particularly if you are only able to get a single turn (e.g. wire simply passing through) of the conductor through the device. 

A common material for snap-on chokes is "Mix 43" - a relatively inexpensive ferrite that is widely used and has best efficacy above 10 MHz.  Having a permeability of around 800, it will reasonably add a bit of extra reactance (resistance to RF) to the conductor over which it is installed.  In described case, I used two devices:
  • Fair-Rite 0443800506, which has an inside diameter of 13.2mm (0.52") and is 15.6mm (0.61") long.
  • Fair-Rite 0443806406, which has an inside diameter of 15.5mm (0.61") and is 16.2mm (0.63") long.
Note:  The above devices - and those below - are available from Mouser Electronics.

 The former will fit nicely over standard "12/2" "Romex" cable while the latter will fit over two of these cables:  The larger '6406 is what is used in Figure 4 right above the box's clamp and in Figure 5 over which three turns of the green electrical cord are wound while the upper two in Figure 4 are the '0506.

As mentioned before, these devices are typically used above 10 MHz and because of their nature, their efficacy decreasing with frequency.  Taking the '0406, it's rated impedance at 10 MHz is 24 ohms and 43 ohms at 25+ MHz, implying an inductance of approximately 0.33 microHenries.  To be sure, this isn't very much, and scaling this for other frequencies implies that at 40 meters its impedance would be on the order of 15 ohms, around 7 ohms at 80 meters and around 4 ohms at 160 meters.  Because the inductance increases with the square of the number of turns, the three turns wound around the green cable (in figure 5) considerably increased its efficacy - probably into the 30-50 ohm range on 40 meters.

This doesn't sound like much - and it isn't - but it is often enough to add just a bit of reactance to the connecting cables to reduce the amount of RF current flowing through the device - which is probably fairly well-protected in its own right - and also to slightly shift self-resonant frequencies in the wiring and the device to reduce the amount of intercepted RF.

What if it hadn't worked?

In my case, the Mix 43 devices were chosen because they were comparatively inexpensive, I had quite a few on hand and they did the job.  If the had not done the job, I would have used different devices - ones with a higher permeability so that the impedance of the conductor(s) over which it was used would also be higher.

The next logical step is to use Mix 31 which has nearly twice the permeability and, for whatever reason, are typically much longer in length.  Take, as an example, two devices of approximately the same inside diameter (e.g. able to accommodate one or two cables) as the two mentioned above:
  • Fair-Rite 0431164181 - 13.05mm I.D. (0.51") and about 31mm (1.22") long.
  • Fair-Rite 0431173551 - 18.8mm I.D. (0.77") and about 42mm (1.65") long.
Because of the size and length and material, these offer about twice the reactance as shorter versions - and cost about four times as much, too!

Even more impedance:

For those really difficult situations, there's yet another material - Mix 75 - and we can get devices similar to those above in that material as well:
  • Fair-Rite 475164181 - 13.05mm I.D. (0.51") and about 39mm (1.54") long.
  • Fair-Rite 475176451 - 18.7mm I.D. (0.74") and about 47mm (1.85") long.
Mix 75 isn't as effective at VHF, but that's not usually where the problem with RF susceptibility usually occurs.  They are about the best thing for HF operation, but at a cost - literally:  Because of the material - and the larger size of the above devices - they tend to cost nearly ten times as much as the Mix 43 material, but if you can only manage to get a single turn of a conductor through it (as would be the case for existing, in-situ wiring) it's the best thing to use.

For example, the Mix 31 '6406 device mention above provides only about 24 ohms of impedance at 10 MHz, but the Mix 75 '4181 device above has 100 ohms at the same frequency and close to 80 ohms at 40 meters.  To be sure, it's a bit of apples and oranges comparison because the '4181 device is about twice as long.

Final comments:

Similar techniques should work on AFCIs (Arc Fault Circuit Interruptors) as well as smoke/fire alarm cables that connect units together - just make sure everything still works when you are done!

In my case, the "semi-unfinished" state of the room allowed access to the "Romex" wires to the outlet, but this is likely not the case when wiring is concealed inside the wall - often with no additional slack in the conductor to permit installation of a ferrite.  If this is the case, the obvious first step toward diagnosing the issue would be to unplug anything on that circuit to see if it still tripped.

If the protector still trips with nothing plugged in, you'll have to get creative, perhaps replacing the box with a larger version (one that is deeper and/or has another "gang" position on it) or even adding a nearby box (with a blank cover plate) to which the wire could be pulled back and ferrite devices installed.

Whatever you do, be safe and sensible and make sure that everything still works as it should when  you are done!


This page stolen from